《Table 4 Recognition accuracy of different algorithms》

  提示：宽带有限、当前游客访问压缩模式

本系列图表出处文件名：随高清版一同展现

《Individual Dairy Cattle Recognition Based on Deep Convolutional Neural Network》

1. 获取 高清版本忘记账户？点击这里登录

1. 下载图表忘记账户？点击这里登录

Basing on the images pretreatment and the parameters discussion，w e conduct comparative experiments w ith SIFT and BOF algorithms in order to verify the validity of algorithm in this paper.During the experiment，the netw ork includes 9layers，the convolution kernel size is 9×9 and the dimension of feature maps in full connection layer is 4 096.BOF algorithm uses speed-up robust features（SURF）as feature descriptor.Different algorithms are adopted to identify the dairy cattle individuals’category for 10，15，and 20.The results of recognition are show n in Table 4.The average recognition accuracy of the algorithm in this paper reaches 96.8%.Compared w ith the traditional methods it has improved significantly，6.2%higher than SIFT algorithm in Ref.[6]and 2.3%higher than BOF algorithm in Ref.[7].